August 2026 links
Although AI increases the number of pro se cases, how does it affect the probability of such defendants winning those cases?
The risk of heat-related death is highest in Europe, despite not being the warmest continent. Our opposition to AC is killing us.
"Under weak conditions, the growth rate of the value of a statistical life (VSL), together with standard Euler-equation objects, identifies the growth rate of lifetime utility". We tradeoff consumption and mortality risk, and the implied discount rates proxy the value of remaining life. In the US, “lifetime utility may have risen by more than a factor of five since 1940, whereas conventional consumption-based calculations using a stable log/CRRA flow utility imply much smaller gains”. We can indirectly measure the extent to which GDP estimates diverge from living standards. Nonetheless, growth matters.
Fair to say that degrowth is a failure.
How “social value” becomes a racket for left-wing propaganda and DEI in policy. I see this in many local galleries, where the promotion of pet causes (such as nuclear disarnament or “queer” issues) the overall theme to be overtly leftist. Of course, many of the visitors are the same leftists currently voting for this government, so redistribution of nonrival goods (such as left-themed galleries) buys political support. Given education polarisation, our elite institutions also end up being dominated by the left, which further increases the government's payoff from investing in left-wing causes. Growth suffers as a result.
Good point. Why should the most productive regions be forced to subsidise (or "invest" in) the least productive?
The current US administration's approach to AI regulation slows such diffusion in political power, whilst disproportionately benefiting rent-seekers. Anthropic was correct that AI needed to be treated as a top national security issue, yet by no means is this equilibrium optimal, especially when we already have CAISI for this very purpose. Ironically, cybersecurity itself may suffer.
For some rare earths that military applications rely on, China holds a virtual monopoly over supply. If this isn't a national security risk, then I don’t know what is, and these inputs aren't really substitutable, so production really is Leontief for uses of these materials. If tariffs reduce our dependence, then the optimal tariff is positive for some goods. Whether this works out empirically is still one for the jury. However, if everyone behaves like this (or some players retaliate), then we have a classic “prisoner's dilemma”.
Beware that measurement invariance does not hold for test scores. This could also be contributing towards the observable magnitude of the Flynn effect, although I still believe such to be non-zero (or else all investment into human capital yields zero return).
Updates on British polling. Race is becoming less predictive of political orientation. The left-right divide is determined mainly by age and education. The populist-mainstream divide (Reform, Restore, Greens, independents vs Labour, Tories, and LDs) is driven by income.
A lot of synaptic pruning characterises maturity.
Wildly different estimates of cosmic dipole anisotropy should leave us sceptical of the notion that FLRW is violated. These recent observations might make great New Scientist articles, yet they're not robust and we cannot rule out measurement error.
I wonder to what extent climate hysteria has contributed to the fertility decline? If there is an effect, then those propagating this have blood on their hands, and cost more lives than slowing climate change would save.
“More licensing is related with lower GDP, worse governance, and more informality”.
"It might not be particularly reassuring, however, to hear that denkiburo are unregulated". This is how free societies regulate market activity: mostly via consumer preferences, which supplies accountability.
In other words, basically gradient descent. LLMs have their value functions specified explicitly by a kernel or via repeated reward fine-tuning. Ours are encoded via centuries of evolution. The intuition, that what we consider consciousness is optimisation, holds. I've never seen a clear definition of consciousness anyhow. Is it cognition and information processing? Self-awareness? Emotion? Optimising relative to some goal (e.g. reproduction)?
How much of the fertility decline reflects a market failing to clear?
On the supply side at least, wokeness is a function of the Friedman doctrine?
The Industrial Revolution were precursors to communism and the two World Wars in many countries, as a result of the socioeconomic changes of the time. Yet this was no reason to reject electricity and the steam engine. So why should we reject AI? If AI is a change on the scale of the Industrial Revolution though, be prepared for geopolitical conflict and new crimes against humanity (bioweapons and cybersecurity are particular concerns). Be prepared for some countries to fall behind the economic frontier at least to the extent as occured since industrialisation, with gaps at least as large as the first to third world gap. Yet be prepared for achievements at least on the scale as the gradual erosion in extreme poverty and child mortality, the airplane, TV, antibiotics, and the gradual elimination of heart disease and obesity.
The public choice theory of the next eschatological hysteria.
Are we just going to sit back and let our government do this to our country?
Indeed. Note that Chinese open source was used by Hugging Face in its cyber defence. Perhaps we may come to rely on them.
Compelling thesis that we’ve underestimated growth and misspecified its dynamics.
We’re going to witness the birth of a new industry of prediction market bets on maths problems. Arguably, AI's greatest strengths lie in mathematics. This implies cybersecurity is probably the field that will change the most, for better or worse.
There are many things the UK screws up on in terms of policy, yet we do animal welfare better than most countries. This is one of the defining moral crises of our times.
Great analysis on how AI has affected employment and labour production functions thus far.
In many aspects, the Tories might be up there as amongst the most leftwing governments in British history.
Criminalising euthanasia is a textbook example of repugnance constraining markets, individual liberty, and our agency. Alleviating suffering is hampered as a result.
Is AI 2027 correct, both positively and in many respects on the normative goals?
Yes the consciousness debate matters as it bears direct implications on how we govern AI. Soon we might want to grant autonomous AI agents legal personhood (or some concept mirroring incorporation) in order to hold them liable for sanctioned acts. Functionalism gets us to this the closest.
Why preserve books when they carry value literally only because of the text within them?